The increasing operating speeds and computing power of microelectronic devices have recently given rise to the need for an increase in the complexity and functionality of the semiconductor structures from which that these devices are fabricated. Hetero-integration of dissimilar semiconductor materials, for example, III-V materials, such as gallium arsenide, gallium nitride, indium aluminum arsenide, and/or germanium with silicon or silicon-germanium substrate, is an attractive path to increasing the functionality and performance of the CMOS platform. In particular, heteroepitaxial growth can be used to fabricate many modern semiconductor devices where lattice-matched substrates are not commercially available or to potentially achieve monolithic integration with silicon microelectronics. Performance and, ultimately, the utility of devices fabricated using a combination of dissimilar semiconductor materials, however, depends on the quality of the resulting structure. Specifically, a low level of dislocation defects is important in a wide variety of semiconductor devices and processes, because dislocation defects partition an otherwise monolithic crystal structure and introduce unwanted and abrupt changes in electrical and optical properties, which, in turn, results in poor material quality and limited performance. In addition, the threading dislocation segments can degrade physical properties of the device material and can lead to premature device failure.
As mentioned above, dislocation defects typically arise in efforts to epitaxially grow one kind of crystalline material on a substrate of a different kind of material—often referred to as “heterostructure”—due to different crystalline lattice sizes of the two materials. This lattice mismatch between the starting substrate and subsequent layer(s) creates stress during material deposition that generates dislocation defects in the semiconductor structure.
Misfit dislocations form at the mismatched interface to relieve the misfit strain. Many misfit dislocations have vertical components, termed “threading segments,” which terminate at the surface. These threading segments continue through all semiconductor layers subsequently added to the heterostructure. In addition, dislocation defects can arise in the epitaxial growth of the same material as the underlying substrate where the substrate itself contains dislocations. Some of the dislocations replicate as threading dislocations in the epitaxially grown material. Other kinds of dislocation defects include stacking faults, twin boundaries, and anti-phase boundaries. Such dislocations in the active regions of semiconductor devices, such as diodes, lasers and transistors, may significantly degrade performance.
To minimize formation of dislocations and associated performance issues, many semiconductor heterostructure devices known in the art have been limited to semiconductor layers that have very closely—e.g. within 0.1%—lattice-matched crystal structures. In such devices a thin layer is epitaxially grown on a mildly lattice-mismatched substrate. As long as the thickness of the epitaxial layer is kept below a critical thickness for defect formation, the substrate acts as a template for growth of the epitaxial layer, which elastically conforms to the substrate template. While lattice matching and near matching eliminate dislocations in a number of structures, there are relatively few lattice-matched systems with large energy band offsets, limiting the design options for new devices.
Accordingly, there is considerable interest in heterostructure devices involving greater epitaxial layer thickness and greater lattice misfit than known approaches would allow. For example, it has long been recognized that gallium arsenide grown on silicon substrates would permit a variety of new optoelectronic devices marrying the electronic processing technology of silicon VLSI circuits with the optical component technology available in gallium arsenide. See, for example, Choi et al, “Monolithic Integration of Si MOSFET's and GaAs MESFET's”, IEEE Electron Device Letters, Vol. EDL-7, No. 4, April 1986. Highly advantageous results of such a combination include high-speed gallium arsenide circuits combined with complex silicon VLSI circuits, and gallium arsenide optoelectronic interface units to replace wire interconnects between silicon VLSI circuits. Progress has been made in integrating gallium arsenide and silicon devices. See, for example, Choi et al, “Monolithic Integration of GaAs/AlGaAs Double-Heterostructure LED's and Si MOSFET's” IEEE Electron Device Letters, Vol. EDL-7, No. 9, September 1986; Shichijo et al, “Co-Integration of GaAs MESFET and Si CMOS Circuits”, IEEE Electron Device Letters, Vol. 9, No. 9, September 1988. However, despite the widely recognized potential advantages of such combined structures and substantial efforts to develop them, their practical utility has been limited by high defect densities in gallium arsenide layers grown on silicon substrates. See, for example, Choi et al, “Monolithic Integration of GaAs/AlGaAs LED and Si Driver Circuit”, IEEE Electron Device Letters, Vol. 9, No. 10, October 1988 (p. 513). Thus, while basic techniques are known for integrating gallium arsenide and silicon devices, there exists a need for producing gallium arsenide layers having a low density of dislocation defects.
To control dislocation densities in highly-mismatched deposited layers, there are three known techniques: wafer bonding of dissimilar materials, substrate patterning, and composition grading. Bonding of two different semiconductors may yield satisfactory material quality. Due to the limited availability and high cost of large size Ge or III-V wafers, however, the approach may not be practical.
Techniques involving substrate patterning exploit the fact that the threading dislocations are constrained by geometry, i.e. that a dislocation cannot end in a crystal. If the free edge is brought closer to another free edge by patterning the substrate into smaller growth areas, then it is possible to reduce threading dislocation densities. In the past, a combination of substrate patterning and epitaxial lateral overgrowth (“ELO”) techniques was demonstrated to greatly reduce defect densities in gallium nitride device, leading to fabrication of laser diodes with extended lifetimes. This process substantially eliminates defects in ELO regions but highly defective seed windows remain, necessitating repetition of the lithography and epitaxial steps to eliminate all defects. In a similar approach, pendeo-epitaxy eliminates substantially all defects in the epitaxial region proximate to the substrate but requires one lithography and two epitaxial growth steps. Furthermore, both techniques require the increased lateral growth rate of gallium nitride, which has not been demonstrated in all heteroepitaxial systems. Thus, a general defect-reduction process utilizing a minimum of lithography/epitaxy steps that does not rely on increased lateral growth rates would be advantageous both to reduce process complexity and facilitate applicability to various materials systems.
Another known technique termed “epitaxial necking” was demonstrated in connection with fabricating a Ge-on-Si heterostructure by Langdo et al. in “High Quality Ge on Si by Epitaxial Necking,” Applied Physics Letters, Vol. 76, No. 25, April 2000. This approach offers process simplicity by utilizing a combination of selective epitaxial growth and defect crystallography to force defects to the sidewall of the opening in the patterning mask, without relying on increased lateral growth rates. Specifically, as shown in FIGS. 1A and 1B, in the (111)<110> diamond cubic slip system, misfit dislocations lie along <110> directions in the (100) growth plane while the threading segments rise up on (111) planes in <110> directions. Threading segments in <110> directions on the (111) plane propagate at a 45° angle to the underlying Si (100) substrate surface. Thus, if the aspect ratio of the holes in the patterning mask is greater than 1, threading segments will be blocked by the mask sidewall, resulting in low-defect top Ge “nodules” formed directly on Si. One important limitation of epitaxial necking, however, is the size of the area to which it applies. In general, as discussed in more detail below, the lateral dimensions (designated as l in FIG. 1A) in both dimensions have to be relatively small in order for the dislocations to terminate at sidewalls.
Thus, there is a need in the art for versatile and efficient methods of fabricating semiconductor heterostructures that would constrain dislocation defects in a variety of lattice-mismatched materials systems. There is also a need in the art for semiconductor devices utilizing a combination of integrated lattice-mismatched materials with reduced levels of dislocation defects for improved functionality and performance.